In Vitro Comparison of Trueness and Precision of an AI-Driven Real-Time Library Matching Protocol with Irregular Geometry Scan Bodies for Full-Arch Implant Scanning

Background: Accurate digital transfer of implant positions is critical for the long-term success of full-arch prosthetic rehabilitation. Photogrammetry remains the benchmark for accuracy, but its high cost and complexity limit clinical adoption. Artificial intelligence (AI)-driven intraoral scanning protocols incorporating real-time library matching and irregular, individually coded scan bodies have been proposed as accessible alternatives to improve accuracy and reproducibility. Methods: This in vitro study evaluated the trueness and precision of a full-arch implant scanning workflow using an AI-assisted real-time library matching system in combination with irregular multi-geometry titanium scan bodies. A high-accuracy structured-light scanner served as the reference standard. Six implant positions (35, 33, 31, 41, 43, 45) were scanned across 20 datasets (n = 120). Mean surface deviations were calculated against the reference STL using CloudCompare v.2.14. and a two-way ANOVA (α = 0.05) in SPSS tested the effects of implant position and scan iteration. Results: The workflow achieved a mean deviation of 13.55 ± 9.70 μm (range 0.77–43.46 μm) across all positions. Anterior sites showed the lowest deviations (e.g., position 31: 3.95 μm; 45: 5.96 μm), while posterior sites exhibited higher deviations (e.g., position 43: 26.15 μm). No mean deviation exceeded 30 μm, and no individual measurement surpassed 45 μm. Implant position significantly affected accuracy (p < 0.001), whereas scan iteration did not (p > 0.05). Conclusions: Within the limitations of this in vitro model, an AI-assisted real-time library matching workflow used in conjunction with irregular multi-geometry scan bodies achieved accuracy levels well within clinically acceptable ranges for full-arch implant impressions. Although comparable to values reported for photogrammetry under laboratory conditions, clinical equivalence should not be assumed. Further in vivo validation is required to confirm performance under routine clinical conditions.